|Publication number||US6511887 B1|
|Application number||US 09/596,747|
|Publication date||Jan 28, 2003|
|Filing date||Jun 19, 2000|
|Priority date||Jun 19, 2000|
|Publication number||09596747, 596747, US 6511887 B1, US 6511887B1, US-B1-6511887, US6511887 B1, US6511887B1|
|Inventors||Mo-Chiun Yu, Syun-Ming Jang|
|Original Assignee||Taiwan Semiconductor Manufacturing Computer|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (7), Referenced by (10), Classifications (7), Legal Events (4)|
|External Links: USPTO, USPTO Assignment, Espacenet|
(1) Field of the Invention
This invention relates to semiconductor integrated circuit devices, and more particularly to a method for making field effect transistors (FETs) with different gate oxide thicknesses for various integrated circuit applications, such as forming a thicker tunnel oxide for non-volatile (flash) memory while providing a thinner gate oxide for high-performance logic circuits. The method uses a patterned thin silicon nitride (Si3N4) layer and a single oxidation step to provide a good quality thicker FET tunnel gate oxide while concurrently forming a good quality thinner FET gate oxide for the high-speed CMOS logic.
(2) Description of the Prior Art
Merged semiconductor logic and memory circuits are finding extensive use in the electronics industry for merged logic/memory devices. These deep submicrometer circuits require different gate oxide thicknesses to optimize the FET performance. Typically flash (non-volatile) memory and peripheral input-output (I/O) circuits require a thicker gate oxide, while CMOS logic circuits require a thinner gate oxide for increased performance (switching speed). For example, flash memory and peripheral circuits require a thicker oxide of 50 Angstroms or more and operate at a gate voltage of about 3.0 to 5.0 volts, while CMOS logic devices have effective gate oxides that are less than 25 Angstroms and operate at lower gate voltages (V) of between about 1.8 V and 2.5 V to achieve higher switching speeds.
One prior art method of achieving dual-thickness gate oxides for FETs is depicted in the schematic cross-sectional views of FIGS. 1 and 2. A field oxide 12 is formed in and on the silicon substrate 10 using conventional means to surround and electrically isolate the logic device areas L, as shown in the right potion of FIG. 1, and as shown in the left portion of FIG. 1 for the memory and peripheral device areas M. A first gate oxide 13 is grown on both the logic and memory/peripheral device areas. Next, as shown in FIG. 2, a photoresist mask 16 is used to mask the gate oxide 13 over the memory device areas M, and then a wet etch is used to remove the gate oxide 13 over the logic device areas L. The photoresist mask 16 is then removed by plasma ashing and/or by stripping in a photoresist stripper. Unfortunately, the photoresist 16 in direct contact with the thin gate oxide 13 on the silicon substrate 10 surface results in contamination that causes reliability problems. Next as shown in FIG. 3, prior to growing a thinner second gate oxide 15 on the logic device areas L, it is necessary to remove the native oxide in the logic device areas using a wet etch. However, the exposure of the first gate oxide 13 to the wet etch makes it difficult to accurately control the thickness of the thicker gate oxide, composed of oxide layers 13 and 15, over the memory device areas when the second oxidation step is performed.
Numerous methods of forming logic circuits with embedded memory devices having dual gate oxides have been reported in the literature. One method of making a dual-thickness gate oxide is described in U.S. Pat. No. 5,960,289 to Tsui et al. In Tsui's method a first gate oxide is deposited and a protection layer (Si3N4) is deposited. The two layers are then patterned to expose some of the device areas. The exposed device areas are oxidized to form a thin second gate oxide for high-performance circuits, while the first gate oxide and the protection layer serve as the thicker gate dielectric for the DRAM and the SRAM cells.
Sugaya in U.S. Pat. No. 4,945,068, teaches a method for forming a thick and thin dual-thickness gate oxide using only a single oxidation step. The method achieves the dual thickness by using an ion implant mask and implanting (injecting) nitrogen (N2) ions into some of the silicon substrate device areas while masking from implant other device areas. A thermal oxidation is performed in which the oxide grows slower on the N2 implanted areas than in non-implanted device areas.
Ong in U.S. Pat. No. 5,880,041 describes a method for using high-pressure oxidation to grow a gate oxide on a silicon substrate. The high pressure increases the oxidation rate (Angstroms/minute) and decreases the oxidation time, which reduces the thermal budget.
Fang et al. in U.S. Pat. No. 5,668,035 describe a method for forming a dual gate oxide for memory with embedded logic technology. The method grows a first gate oxide over memory and logic device areas. Then a polysilicon layer is deposited and patterned to expose the first gate oxide over the logic device areas and to protect the first gate oxide over the memory device areas. The first gate oxide is removed by wet etch over the logic device areas and a thinner second gate oxide is grown over the logic device areas.
Matsuoka et al. in U.S. Pat. No. 5,926,741 describe a method for forming a gate dielectric layer without generation of a natural oxide film. The method involves cleaning the silicon wafer in an inert gas ambient and replacing the ambient gas with an oxidizing gas and applying heat to form the gate oxide on the wafer.
However, there is still a need in the semiconductor industry for fabricating dual-thickness gate oxides using a simple cost-effective process, while preventing native oxides from forming on device areas, as well as avoiding boron penetration in the thin gate oxides for P-channel FETs.
It is a principal object of the present invention to form dual gate oxides on a single integrated circuit chip for field effect transistors (FETs) for making a thin effective gate oxide for high-performance logic circuits and a thicker gate oxide for non-volatile memory and/or peripheral circuits.
It is another object of the invention to use a patterned thin silicon nitride (Si3N4) layer to form the thin gate oxide while concurrently forming the thicker gate oxide using a single oxidation step.
A further objective of this invention is to use a single oxidation step to reduce processing complexity and to improve product reliability.
The method for making this dual gate oxide using a patterned thin Si3N4 layer for CMOS logic circuits and for non-volatile memory is now briefly described. The method begins by providing a semiconductor substrate. The substrate is typically a single-crystal silicon having a <100> crystallographic orientation. A field oxide is formed surrounding and electrically isolating CMOS logic device areas and memory device areas. Typically for these more advanced high-density integrated circuits, the field oxide is shallow trench isolation as commonly practiced in the industry. By the method of this invention the dual gate oxides are now formed by removing any native oxide that has inadvertently formed on the logic and memory device areas. Next a very thin blanket Si3N4 layer (e.g., 10-20 Angstroms) is deposited on the substrate using chemical vapor deposition (CVD). Preferably the native oxide is removed in situ just prior to depositing the Si3N4 layer to avoid reforming a native oxide. The Si3N4 layer is then patterned using a photoresist mask and a hot phosphoric acid (H3PO4) etch. The Si3N4 is patterned to leave portions over the logic device areas while exposing the surface of the memory and/or peripheral device areas. A key feature of this invention is that the photoresist mask used to pattern the Si3N4 does not come in direct contact with a gate oxide, as in the prior art. Therefore the gate oxide and/or the substrate are not contaminated by the photoresist. Any second native oxide that has inadvertently formed on the memory device areas is now removed in a dilute hydrofluoric (HF) acid solution or an HF vapor etch, while essentially leaving unetched the patterned Si3N4 over the logic device areas. The surface of the substrate is then subjected to a single oxidation step to form a thicker gate oxide (greater than or equal to 50 Angstroms) on the memory device areas, while concurrently converting the thin patterned Si3N4 layer to a nitrogen-rich SiO2 layer having an effective oxide thickness of less than or equal to about 25 Angstroms over the logic device areas. The oxidation is carried out for a time sufficient to convert the Si3N4 to a nitrogen-rich SiO2 and to partially oxidize the underlying surface of the silicon substrate to form a good interface. After completing this improved dual gate oxide, conventional processing is used to form FET devices. For example, the FET gate electrodes can be formed next by depositing and patterning a doped polysilicon layer or a polycide layer.
FIGS. 1-3 show schematic cross-sectional views of a method for forming a thick and thin dual gate oxide by the prior art showing the limitations for controlling oxide thickness and photoresist contamination.
FIGS. 4 through 7 show schematic cross-sectional views for making a thick and thin dual gate oxide using a thin Si3N4 layer and a single oxidation step by the method of this invention.
The method for making dual gate oxides having different thicknesses for deep sub-micron CMOS integrated circuits is now described in detail. The method uses a patterned thin Si3N4 layer and a single oxidation step to achieve the dual thicknesses of the gate oxides. Although the method is shown for making gate oxides for N channel FETs on a P substrate to simplify the discussion and drawings, it should also be understood by one skilled in the art that by including additional process steps, for example, by including N-well and P-well regions on a substrate, P-channel and N-channel FETs can be made from which Complementary Metal-Oxide-Semiconductor (CMOS) circuits can be formed. It should also be understood that while providing a thin gate oxide for improving the switching speed of the logic circuits, a thicker gate oxide is provided for improving higher voltage FET devices for peripheral input/output (I/O) buffer circuits, for flash (non-volatile) memory and the like on the chip.
Referring now to FIG. 4, the method begins by providing a substrate 10 having logic regions L for logic circuits and memory regions M which represent device areas for flash memory and the like. The memory regions are also representative of the type of gate oxide required for higher voltage I/O peripheral circuits. Typically for N-channel FETS, the substrate 10 is a P doped single-crystal silicon having a <100> crystallographic orientation. Field oxide regions 12 are formed surrounding and electrically isolating the device areas in both the logic regions L and the memory regions M. For advanced high-density circuits the preferred field oxide 12 is a Shallow Trench Isolation (STI), as is commonly practiced in the semiconductor industry. Generally the STI is formed by etching trenches in the field oxide areas on the substrate to a depth of between about 3000 and 5000 Angstroms. The trenches are filled with an insulating material, such as SiO2, that is made planar with the surface of the substrate 10 using a planarizing etching back or by chemical-mechanical polishing (CMP).
Still referring to FIG. 4, and more specific to the method of this invention, the gate oxide having dual oxide thicknesses are now formed as now described. Any native oxide (not shown in the Fig.) that has inadvertently formed on the logic and memory device areas is removed. The preferred method for removing this first native oxide is by a hydrofluoric (HF) acid vapor etch in situ in the deposition tool to minimize the re-formation of a native oxide prior to depositing a Si3N4 layer. Alternatively the native oxide can be removed in an HF solution, typically made up of 1:100 HF to H2O by volume, just prior to loading the substrate (wafer) into a Si3N4 deposition tool. Next a thin blanket Si3N4 layer 14 is deposited on the substrate 10. The Si3N4 layer 14 is deposited by chemical vapor deposition using, for example, dichlorosilane (SiCl2H2) and ammonia (NH3) as the reactant gas mixture. Si3N4 layer 14 is deposited to a preferred thickness of between about 10 and 20 Angstroms.
Referring now to FIG. 5, the Si3N4, layer 14 is patterned using a photoresist mask 16 and a wet etch to remove the Si3N4 over the device areas in the memory regions M, while protecting from etching the Si3N4 over the device areas in the logic regions L. Preferably the etching is carried out using a hot phosphoric acid (H3PO4) etchant, which prevents substrate damage, and the etching is typically carried out at a temperature of between about 100 and 200° C. A key feature of this invention is that the Si3N4 layer 14 prevents mobile ions from diffusing and protects the substrate from photoresist contamination during photoresist processing and etching. Also the gate oxide is not affected since it has not yet been formed on the silicon substrate 10, as in the prior art.
Referring to FIG. 6, the photoresist 16 shown in FIG. 5 is stripped by plasma ashing in O2 and/or wet stripping. Any second native oxide (not shown) that may inadvertently have formed on the exposed memory device regions M is now removed. The second native oxide is removed similar to the removal of the first native oxide using a HF acid vapor etch in situ in the oxidation chamber (furnace) to minimize the re-formation of a native oxide. Alternatively any second native oxide can be removed in an HF solution, typically made up of 1:100 HF to H2O by volume, just prior to loading the substrate (wafer) into the oxidation chamber, while essentially leaving unetched the patterned Si3N4 layer 14 over the logic device areas L.
Referring to FIG. 7, the surface of the substrate 10 is then subjected to a single oxidation step to form a thicker gate oxide layer 18 (greater than or equal to 50 Angstroms) on the memory device areas M, while concurrently slowly converting the thin patterned Si3N4 layer 14 to a nitrogen-rich SiO2 layer 14′ having an effective oxide thickness (EOT) of less than or equal to about 25 Angstroms over the logic device areas L. The oxidation is carried out for a time sufficient to convert the Si3N4 to a nitrogen-rich SiO2, commonly referred to as silicon oxynitride (SiON). The oxidation step is carried out for a sufficient time to partially oxidize the underlying surface of the silicon substrate 10 to form a good interface. The oxidation is preferably carried out in an oxidation furnace using a wet oxidation (steam oxidation). The preferred method is to perform a rapid thermal anneal at a temperature greater than 900° C. and for a time less than 3 minutes. This temperature and time can be adjusted to achieve the desired effective oxide thickness (EOT) of the SiON 14′ for the desired electrical parameters. The nitrogen-rich SiON layer 14′ also prevents boron diffusion on P-channel PET devices to improve electrical stability. After completing this novel, improved dual gate oxide, conventional processing can be used to form the FETs for the logic and memory devices. For example, the FET gate electrodes 20 can be formed next by depositing and patterning a doped polysilicon layer or a polycide layer.
While the invention has been particularly shown and described with reference to the preferred embodiment thereof, it will be understood by those skilled in the art that various changes in form and details may be made without departing from the spirit and scope of the invention.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US4945068||Oct 25, 1989||Jul 31, 1990||Matsushita Electronics Corporation||Manufacturing method of semiconductor nonvolatile memory device|
|US5502009 *||Feb 16, 1995||Mar 26, 1996||United Microelectronics Corp.||Method for fabricating gate oxide layers of different thicknesses|
|US5668035||Jun 10, 1996||Sep 16, 1997||Taiwan Semiconductor Manufacturing Company Ltd.||Method for fabricating a dual-gate dielectric module for memory with embedded logic technology|
|US5880041||May 27, 1994||Mar 9, 1999||Motorola Inc.||Method for forming a dielectric layer using high pressure|
|US5926741||Jul 7, 1997||Jul 20, 1999||Sharp Kabushiki Kaisha||Method of forming gate dielectric films for MOSFETs without generation of natural oxide films|
|US5960289 *||Jun 22, 1998||Sep 28, 1999||Motorola, Inc.||Method for making a dual-thickness gate oxide layer using a nitride/oxide composite region|
|US6033958 *||Jun 30, 1998||Mar 7, 2000||United Microelectronics Corp.||Method of fabricating dual voltage MOS transistors|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US6642117 *||Aug 5, 2002||Nov 4, 2003||Taiwan Semiconductor Manufacturing Co. Ltd||Method for forming composite dielectric layer|
|US6670226 *||Mar 8, 2002||Dec 30, 2003||Taiwan Semiconductor Manufacturing Co., Ltd||Planarizing method for fabricating gate electrodes|
|US6709932 *||Aug 30, 2002||Mar 23, 2004||Texas Instruments Incorporated||Method for improving gate oxide integrity and interface quality in a multi-gate oxidation process|
|US7525156||Jun 8, 2007||Apr 28, 2009||International Business Machines Corporation||Shallow trench isolation fill by liquid phase deposition of SiO2|
|US7776708 *||Aug 11, 2005||Aug 17, 2010||National Semiconductor Corporation||System and method for providing a nitride cap over a polysilicon filled trench to prevent formation of a vertical bird's beak structure in the manufacture of a semiconductor device|
|US7863153||Jul 13, 2006||Jan 4, 2011||National Semiconductor Corporation||System and method for creating different field oxide profiles in a locos process|
|US7932152 *||Apr 26, 2011||Chartered Semiconductor Manufacturing, Ltd.||Method of forming a gate stack structure|
|US8361863||Jan 29, 2013||Mosys, Inc.||Embedded DRAM with multiple gate oxide thicknesses|
|US20050189598 *||Apr 16, 2004||Sep 1, 2005||Taiwan Semiconductor Manufacturing Co., Ltd.||Logic embedded-memory integrated circuits|
|US20090197387 *||Feb 5, 2008||Aug 6, 2009||Chartered Semiconductor Manufacturing, Ltd.||Method of forming a gate stack structure|
|U.S. Classification||438/275, 257/E21.625, 438/221, 438/225|
|Jun 19, 2000||AS||Assignment|
Owner name: TAIWAN SEMICONDUCTOR MANUFACTURING COMPANY, TAIWAN
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:YU, MO-CHIUN;JANG, SYUN-MING;REEL/FRAME:010889/0075
Effective date: 20000418
|Jun 30, 2006||FPAY||Fee payment|
Year of fee payment: 4
|Jul 1, 2010||FPAY||Fee payment|
Year of fee payment: 8
|Jul 2, 2014||FPAY||Fee payment|
Year of fee payment: 12